HackerNews Digest

March 04, 2026

Motorola GrapheneOS devices will be bootloader unlockable/relockable

The page is a Mastodon post from the GrapheneOS account referencing the user @[email protected] and stating that “It will fully support using …”. Aside from this brief message, the site displays the standard Mastodon notice that the web application requires JavaScript and offers alternative native apps. An image placeholder is included with alt text “Mastodon”. No further technical details or content are present.
Read full article →
The discussion expresses optimism that Motorola devices could become a major platform for alternative operating systems like GrapheneOS, while seeking information on supported models and device lists. Concerns are raised about security implications of Lenovo’s ownership, potential restrictions in secure environments, and how GrapheneOS will manage Android source‑code release delays and update provisioning. Technical questions focus on the handling of Widevine L1 keys after bootloader relocking. The broader context reflects apprehension about limited consumer demand for privacy‑focused customization amid market influences from large hardware suppliers.
Read all comments →

MacBook Pro with M5 Pro and M5 Max

Apple announced new 14‑ and 16‑inch MacBook Pro models powered by the M5 Pro and M5 Max chips, built on a dual‑die Fusion Architecture with an up‑to‑18‑core CPU (6 super cores + 12 performance cores) and a next‑generation GPU that includes a Neural Accelerator per core. The chips deliver up to 30 % faster CPU performance, 4‑8 × AI speed‑ups versus prior M‑series, 50 % higher graphics throughput, and up to 4 × faster LLM prompt processing. Unified memory bandwidth reaches 307 GB/s (64 GB max) on M5 Pro and 614 GB/s (128 GB max) on M5 Max. SSD read/write speeds double to 14.5 GB/s, with base storage of 1 TB (M5 Pro) or 2 TB (M5 Max). Other hardware includes the N1 chip for Wi‑Fi 7/Bluetooth 6, three Thunderbolt 5 ports, HDMI 8K, SDXC slot, MagSafe 3, a 12 MP Center Stage camera, six‑speaker audio system, and up to 24 h battery life (fast‑charge 50 % in 30 min). macOS Tahoe adds enhanced Spotlight, Apple Intelligence, Live Translation, and expanded Continuity. Pricing starts at US $2,199 (M5 Pro, 14‑inch) and US $3,599 (M5 Max, 14‑inch); availability begins 11 Mar 2026. The device uses 45 % recycled content and 50 % renewable electricity in its supply chain.
Read full article →
Comments show mixed reactions to the new MacBook Pro models. Many highlight the M5 Pro/Max’s architectural changes, higher memory bandwidth, faster AI‑task claims, larger standard storage and improved battery life as notable upgrades. At the same time, users note the limited 128 GB unified memory ceiling, price increases, exclusion of a power adapter, and lack of built‑in cellular as drawbacks. Skepticism appears around the AI performance marketing, the practical impact for local LLM workloads, and the overall incremental nature of the refresh. Several users express a preference to wait for future generations or alternative hardware, while others seek better Linux support and more sustainable upgrade options.
Read all comments →

Lenovo's New ThinkPads Score 10/10 for Repairability

Lenovo’s latest ThinkPad T14 Gen 7 and T16 Gen 5 received a provisional 10/10 repairability rating from iFixit—the first T‑series laptops to achieve the highest score. The rating reflects a design process in which Lenovo collaborated with iFixit from the outset, integrating serviceability considerations alongside performance, reliability, thermal efficiency and form‑factor goals. Key modular components include a near‑tool‑free battery, industry‑standard M.2 SSD, easily replaceable keyboard, LPC AMM2 memory, a detachable cooling fan, and fully modular Thunderbolt ports; the display assembly also swaps with minimal effort. Remaining limitations are the non‑modular Wi‑Fi module, some I/O ports still integrated on the main board, and adhesive‑bound display panels that complicate panel‑level repairs. Lenovo frames the 10/10 score as a new baseline rather than an endpoint, emphasizing ongoing improvements to reduce repair time, tool requirements, and waste. The achievement is notable because ThinkPad T‑series laptops dominate enterprise fleets, making high‑volume, repairable business laptops a potential industry standard.
Read full article →
Comments show strong appreciation for ThinkPad durability, repairability and Linux compatibility, with many users highlighting easy maintenance, long‑term reliability and robust warranty support. Criticism centers on perceived AI‑generated marketing language in reviews, concerns over Lenovo’s firmware security and lack of replaceable components such as Wi‑Fi modules and soldered ports, and rising prices that diminish value compared with competitors like Framework. Opinions also note occasional design flaws—plastic construction, problematic keyboard keys, and limited upgrade paths—while overall sentiment remains favorable toward the brand’s longevity and service ecosystem.
Read all comments →

You can use newline characters in URLs

None
Read full article →
Comments focus on handling unconventional whitespace characters and non‑standard HTTP methods. Several notes stress that tabs, newlines, and vertical tabs can affect HTML attributes, HTTP headers, and filename compatibility, prompting advice to strip such characters and avoid spaces in filenames. There is agreement that the title of the referenced discussion is misleading. Additional observations mention CDNs like Cloudflare blocking unusual HTTP verbs, while a few remarks add humor or unrelated suggestions, keeping the overall tone neutral and informational.
Read all comments →

The largest acidic geyser has been putting on quite a show

Echinus Geyser, located in the Back Basin of Norris Geyser Basin (≈200 m from Steamboat Geyser), is the world’s largest acidic geyser, though its acidity is comparable to orange juice or vinegar. The pool is ~20 m across; red rims result from iron, aluminum, and arsenic, and silica‑coated spiny rocks give the feature its name. Historically, eruptions were rare before 1948, became regular in the 1970s (40–80 min intervals), and sometimes lasted >90 min in the 1980s‑1990s, reaching up to 23 m. Activity declined in the early 2000s. A temperature sensor installed in 2010 recorded 15 eruptions (Oct 2010–Jan 2011). After a quiet period, eruptions returned in Sep 2017, occurring every 2–3 h from 18 Oct to 10 Nov, then ceased except for isolated events (Jan 2018, Jan 2019, Dec 2020). In Feb 2026, surges resumed, with eruptions on 7, 9, 12, 15 Feb and regular 2–5 h intervals from 16 Feb, lasting 2–3 min and reaching 6–10 m. Eruptions correspond to temperature spikes ≈70 °C, while surges show 40–50 °C spikes. Future activity is uncertain; past patterns suggest brief active phases lasting weeks to months.
Read full article →
None
Read all comments →

Claude's Cycles [pdf]

None
Read full article →
The discussion highlights enthusiasm for using reinforcement‑learning‑scaled probability models to make expert problem‑solving more accessible, noting LLMs’ broad knowledge, ability to form connections, and capacity for persistent trial‑and‑error. Examples show human guidance still required, with occasional failures and context limits. Concerns are raised about keeping models current as scientific frontiers expand, questioning continual‑learning versus costly retraining. Opinions also reflect uncertainty about AI’s ability to resolve truly hard problems soon, while observing broader impacts on work, value of human expertise, and potential cultural shifts.
Read all comments →

Voxile: A ray-traced game made in its own engine and programming language

None
Read full article →
The comments blend enthusiasm for the technical accomplishments with cautious appraisal of the game itself. Many express excitement about the Lobster language, the creator’s résumé, and the visual quality, noting a desire to try the game and compare its performance to other engines. At the same time, several points question the practical benefits of voxel‑based physics, criticize the game’s reliance on familiar Minecraft‑style mechanics, and highlight the mixed Steam rating, suggesting that technical novelty does not guarantee compelling gameplay. Overall, the tone is optimistic yet measured.
Read all comments →

Weave – A language aware merge algorithm based on entities

Weave is a Git merge driver that replaces line‑based merging with entity‑level semantic merging using Tree‑Sitter parsers. It parses the base, ours, and theirs versions of a file into semantic entities (functions, classes, JSON keys, etc.), matches entities by name‑type‑scope, and merges each entity independently: unchanged entities are kept, single‑side changes are applied automatically, and concurrent changes to the same entity trigger an intra‑entity 3‑way merge, producing a conflict only when the modifications are truly incompatible. This eliminates false conflicts when unrelated code is added to the same file, a common issue for multiple AI agents or developers. Benchmarks on merge commits from major open‑source repositories show 31/31 clean merges versus Git’s 15/31, with zero regressions and high human‑match rates. Weave supports TypeScript, JavaScript, Python, Go, Rust, JSON, YAML, TOML, Markdown, and falls back to line‑level merging for unsupported or large files. It consists of three components: **weave‑core** (entity extraction and merge algorithm), **weave‑driver** (Git driver binary), and **weave‑cli** (setup and preview commands). Installation involves adding a custom merge driver in `.gitattributes` and configuring Git to invoke Weave for the supported file extensions.
Read full article →
The feedback expresses enthusiasm for the tool, highlighting interest in adding Ruby language support and noting perceived value even without direct experience. The commenter also inquires about the effectiveness of agents within the platform, indicating curiosity about their performance. Overall sentiment is positive and forward‑looking, focusing on desired feature expansion and functional expectations.
Read all comments →

Textadept

Textadept is a fast, minimalist, cross‑platform text editor aimed at programmers. It is implemented in C, C++ and Lua and runs on Windows 10+ (64‑bit/ARM), macOS 13+, Linux (Qt 5/GTK 3 GUI or ncurses terminal) and BSD, delivering self‑contained executables that require no installation. Core features include support for over 100 languages, multiple carets and selections, unlimited split views, highly configurable key bindings (including language‑specific keys, chains and modes), generic and language‑specific nested snippets, and shell‑command integration for building and testing. Nearly every editor component can be scripted or extended with Lua, and the application has no internet connectivity. Pre‑built binaries bundle required runtimes (Qt on Windows/macOS). Optional extra modules can be installed in ~/.textadept/modules and loaded via Lua’s require() in the preferences file. Building from source needs CMake 3.22+, a C/C++ compiler, and at least one UI toolkit (Qt 5.15+, GTK 2.24+, or ncurses). The MIT‑licensed project provides a manual (F1) and Lua API reference (Shift + F1).
Read full article →
The comments express overall approval of TextAdept’s lightweight, fast performance and Qt‑based UI, highlighting its Lua extensibility and open‑source code as strengths. Users note the appeal of its split‑view capability but point out the lack of an automatic temporary‑file save feature and limited documentation on its GUI widget implementation. Comparisons to similar editors such as Geany, Kate, NotepadNext and Lite XL appear, with suggestions to explore Lua APIs for additional functionality while acknowledging the broader value of open‑source text editors.
Read all comments →

Graphics Programming Resources

An overview of variance‑reduction methods for Monte‑Carlo rendering includes: - **Bidirectional Path Tracing** – connects eye and light sub‑paths to capture difficult lighting such as indirect illumination and caustics, improving convergence over unidirectional tracing. - **Metropolis‑Hastings** – employs a Markov‑Chain Monte‑Carlo scheme that mutates sample paths, focusing computation on high‑contribution regions and reducing variance in complex scenes. - **Multiple‑Importance Sampling (MIS)** – combines several sampling strategies with weighted balances to minimize estimator variance across diverse integrands. - **Stratified Sampling** – partitions the domain into strata and draws one sample per stratum, guaranteeing more uniform coverage than pure random sampling. - **Low‑Discrepancy Sequences** – use deterministic quasi‑random point sets (e.g., Halton, Sobol) to achieve faster error decay than stochastic sampling. - **Blue‑Noise Sampling** – generates patterns with minimal low‑frequency power, yielding visually pleasing noise and improved convergence for image‑plane sampling. - **Photon Mapping** – a two‑pass global illumination technique that stores photon contributions in a spatial map, allowing efficient evaluation of indirect lighting. - **Finite‑Element Radiosity** – solves the radiosity integral using discretized surface elements, providing deterministic, low‑variance illumination for diffuse interreflections.
Read full article →
None
Read all comments →